Cerebral hemorrhage follow-up analysis method and device, electronic equipment and storage medium

Abstract

The present application relates to the field of computer, especially to a cerebral hemorrhage follow-up analysis method and device, electronic equipment and computer readable storage medium. The cerebral hemorrhage follow-up analysis method comprises: obtaining each first hemorrhagic lesion and a first position region of the first hemorrhagic lesion from a first cerebral hemorrhage image registered with a second cerebral hemorrhage image; obtaining each second hemorrhagic lesion and a second position region of the second hemorrhagic lesion from the second cerebral hemorrhage image; determining a cerebral hemorrhage type of a patient based on the first position region, the second position region, the first hemorrhagic lesion and the second hemorrhagic lesion, wherein the cerebral hemorrhage type comprises at least one of an expanding or reducing hemorrhage type, a new hemorrhage type and a disappeared hemorrhage type; and determining a display layer using the cerebral hemorrhage type to display the first hemorrhagic lesion and the second hemorrhagic lesion in the display layer. The present application can show the hemorrhagic lesion to the doctor completely at one time, and improve the diagnosis efficiency of the doctor.

Description

Technical Field

[0001] This invention relates to the field of computers, and more particularly to a method, apparatus, electronic device, and computer-readable storage medium for follow-up analysis of cerebral hemorrhage. Background Technology

[0002] Spontaneous intracerebral hemorrhage (CH) refers to the rupture of non-traumatic blood vessels in the brain, leading to blood pooling in the brain parenchyma. Its incidence rate is second only to ischemic stroke among all stroke subtypes. The incidence of CH is (12–15) per 100,000 person-years. In Western countries, CH accounts for approximately 15% of all strokes and 10%–30% of all hospitalized stroke patients. The proportion is even higher in my country, accounting for 18.8%–47.6% of all strokes. CH is a dangerous condition with a mortality rate as high as 35%–52% within 30 days of onset. Only about 20% of patients recover their ability to live independently after 6 months, placing a heavy burden on society and families.

[0003] Follow-up analysis of intracerebral hemorrhage refers to comparing CT images of the hemorrhage patient at different time points (including before and after follow-up) to observe changes in the hemorrhage lesion over time, such as lesion volume and location, thereby assessing the progression of the hemorrhage and guiding clinicians in treatment. Currently, to present the hemorrhage status to doctors, the intracerebral hemorrhage images before and after follow-up are shown layer by layer.

[0004] Since hemorrhage lesions often exist in multiple layers of brain hemorrhage images, the method of displaying each layer one by one cannot present the entire hemorrhage lesion to the doctor at once. The doctor still needs to check each layer one by one and conduct further analysis in order to evaluate the changes in the hemorrhage lesion. Summary of the Invention

[0005] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, one objective of this invention is to propose a method for follow-up analysis of cerebral hemorrhage, which has the advantage of providing doctors with a complete and simultaneous view of the hemorrhage lesion, thereby improving the efficiency of doctors' diagnosis.

[0006] According to a first aspect of the present invention, a method for follow-up analysis of cerebral hemorrhage is provided, comprising:

[0007] Each first hemorrhage lesion and the first location region of the first hemorrhage lesion are obtained from the first brain hemorrhage image after registration with the second brain hemorrhage image. The first brain hemorrhage image is the brain hemorrhage image before follow-up, and the second brain hemorrhage image is the brain hemorrhage image after follow-up.

[0008] Obtain each second hemorrhage lesion and the second location region of the second hemorrhage lesion from the second cerebral hemorrhage image;

[0009] The patient's cerebral hemorrhage type is determined based on the first location region, the second location region, the first hemorrhage lesion, and the second hemorrhage lesion. The cerebral hemorrhage type includes at least one of the following: expanding or decreasing hemorrhage type, newly added hemorrhage type, and disappearing hemorrhage type.

[0010] The type of cerebral hemorrhage is used to determine a display layer, in which the first hemorrhage lesion and the second hemorrhage lesion are displayed.

[0011] In one exemplary embodiment of this disclosure, determining the type of cerebral hemorrhage based on the first location region, the second location region, the first hemorrhage lesion, and the second hemorrhage lesion includes:

[0012] For each of the first hemorrhage lesions, if the hemorrhage lesion type of the first hemorrhage lesion is consistent with the hemorrhage lesion type of the second target hemorrhage lesion in the second hemorrhage lesion, and there is an overlapping area between the first location region and the second target location region of the second target hemorrhage lesion, then the first volume of the first location region and the second volume of the second location region are obtained.

[0013] If the first volume is smaller than the second volume, then the type of cerebral hemorrhage is determined to be an expanding hemorrhage type;

[0014] If the first volume is smaller than the second volume, then the type of cerebral hemorrhage is determined to be a reduced type of hemorrhage.

[0015] In one exemplary embodiment of this disclosure, the method further includes:

[0016] If the bleeding lesion type of the first bleeding lesion is inconsistent with the bleeding lesion type of each of the second bleeding lesions, then the bleeding type is determined to be the disappearance bleeding type.

[0017] In one exemplary embodiment of this disclosure, the method further includes:

[0018] If a second specific bleeding lesion exists within the second bleeding lesion, then the bleeding type is determined to be a newly added bleeding type, and the bleeding lesion type of the second specific bleeding lesion is inconsistent with the bleeding lesion type of each of the first bleeding lesions.

[0019] In one exemplary embodiment of this disclosure, determining the display layer using the type of cerebral hemorrhage includes:

[0020] If the bleeding type is an expanding or decreasing bleeding type, then obtain the union region of the first location region and the second location region;

[0021] Obtain the union region in the first layer of the first brain hemorrhage image / or the second brain hemorrhage image;

[0022] The first layer is used as the display layer.

[0023] In one exemplary embodiment of this disclosure, determining the display layer based on the type of cerebral hemorrhage includes:

[0024] If the hemorrhage type is a newly added hemorrhage type, then obtain the second location region in the second layer of the second brain hemorrhage image;

[0025] The second layer is used as the display layer.

[0026] In one exemplary embodiment of this disclosure, determining the display layer based on the type of cerebral hemorrhage includes:

[0027] If the bleeding type is a disappearing bleeding type, then obtain the third layer of the first location region in the first brain hemorrhage image;

[0028] The third layer is used as the display layer.

[0029] In an exemplary embodiment of this disclosure, before obtaining each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first brain hemorrhage image registered with the second brain hemorrhage image, the method further includes:

[0030] Obtain the deformation field of the first brain hemorrhage image relative to the second brain hemorrhage image;

[0031] The deformation field is used to correct the first cerebral hemorrhage image to obtain the registered first cerebral hemorrhage image.

[0032] In one exemplary embodiment of this disclosure, obtaining the deformation field of the first brain hemorrhage image relative to the second brain hemorrhage image includes:

[0033] Obtain the first brain tissue from the first brain hemorrhage image;

[0034] Second brain tissue was obtained from the second brain hemorrhage image;

[0035] The deformation field is obtained by registering the first brain hemorrhage image and the second brain hemorrhage image using the first brain tissue and the second brain tissue.

[0036] In one exemplary embodiment of this disclosure, the first brain tissue includes a first midline of the brain, and the first brain tissue from which the image of the first cerebral hemorrhage is acquired includes:

[0037] The first brain hemorrhage image is input into a pre-trained brain tissue extraction model to obtain an image of the first falx cerebri region in the first brain hemorrhage image.

[0038] Obtain the middle pixel of each row of pixels in the first falx cerebralis region image;

[0039] The line formed by the intermediate pixels is taken as the first midline of the brain.

[0040] In one exemplary embodiment of this disclosure, obtaining each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first brain hemorrhage image registered with the second brain hemorrhage image includes:

[0041] The registered first cerebral hemorrhage image is normalized to obtain a normalized cerebral hemorrhage image;

[0042] The normalized cerebral hemorrhage image is input into a pre-trained hemorrhage lesion classification model to obtain the first hemorrhage lesion of the registered first cerebral hemorrhage image;

[0043] The first location region of the first hemorrhage lesion in the registered first cerebral hemorrhage image is determined.

[0044] In one exemplary embodiment of this disclosure, normalizing the registered first cerebral hemorrhage image to obtain a normalized cerebral hemorrhage image includes:

[0045] The registered first cerebral hemorrhage image is normalized in size to obtain a normalized cerebral hemorrhage image;

[0046] Determine the region of interest from the size-normalized brain hemorrhage image;

[0047] Pixel normalization is performed on the region of interest to obtain a pixel-normalized brain hemorrhage image.

[0048] In an exemplary embodiment of this disclosure, before obtaining each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first brain hemorrhage image registered with the second brain hemorrhage image, the method further includes:

[0049] Obtain a training set, which includes brain hemorrhage image samples and sample labels corresponding to the brain hemorrhage image samples;

[0050] The brain hemorrhage image samples are augmented to obtain augmented brain hemorrhage image samples;

[0051] The brain hemorrhage image sample is input into the hemorrhage lesion classification model to obtain the sample hemorrhage lesion type of the brain hemorrhage image sample;

[0052] The loss function of the hemorrhage lesion classification model is constructed using the sample hemorrhage lesion type and the label;

[0053] The loss function is used to train the hemorrhage lesion classification model.

[0054] According to a second aspect of this disclosure, a follow-up analysis device for cerebral hemorrhage is provided, comprising:

[0055] The first hemorrhage lesion acquisition module is used to acquire each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first brain hemorrhage image after registration with the second brain hemorrhage image. The first brain hemorrhage image is the brain hemorrhage image before follow-up, and the second brain hemorrhage image is the brain hemorrhage image after follow-up.

[0056] The second hemorrhage lesion acquisition module is used to acquire each second hemorrhage lesion and the second location region of the second hemorrhage lesion from the second brain hemorrhage image;

[0057] A brain hemorrhage type determination module is used to determine the type of brain hemorrhage of a patient based on the first location region, the second location region, the first hemorrhage lesion, and the second hemorrhage lesion. The type of brain hemorrhage includes at least one of the following: expanding or decreasing hemorrhage type, newly added hemorrhage type, and disappearing hemorrhage type.

[0058] The display layer determination module is used to determine a display layer based on the type of cerebral hemorrhage, so as to display the first hemorrhage lesion and the second hemorrhage lesion on the display layer.

[0059] According to a third aspect of this disclosure, an electronic device is provided, comprising:

[0060] processor;

[0061] Memory used to store the processor's executable instructions;

[0062] The processor is configured to execute the instructions to implement the cerebral hemorrhage follow-up analysis method as described in any one of the first aspects.

[0063] According to a fourth aspect of this disclosure, a computer-readable storage medium is provided, wherein when instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the brain hemorrhage follow-up analysis method as described in any one of the first aspects.

[0064] The follow-up analysis method for cerebral hemorrhage provided in this disclosure can determine all display layers of the hemorrhage lesion according to the patient's cerebral hemorrhage type, and display the first hemorrhage lesion and the second hemorrhage lesion on the display layer. It can completely display the hemorrhage lesion to the doctor at one time, thereby improving the doctor's diagnostic efficiency.

[0065] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0066] Figure 1 This is a flowchart of a follow-up analysis method for cerebral hemorrhage according to an exemplary embodiment;

[0067] Figure 2 This is an architecture diagram of a follow-up analysis system for cerebral hemorrhage provided according to an exemplary embodiment;

[0068] Figure 3 This is a block diagram of a follow-up analysis device for cerebral hemorrhage according to an exemplary embodiment;

[0069] Figure 4 This is a schematic diagram of a storage medium provided according to an exemplary embodiment;

[0070] Figure 5 This is a block diagram of an electronic device provided according to an exemplary embodiment. Detailed Implementation

[0071] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0072] The method for follow-up analysis of cerebral hemorrhage according to an embodiment of the present invention will be described below with reference to the accompanying drawings. Figure 1 As shown, the above-mentioned follow-up analysis method for cerebral hemorrhage may include the following steps:

[0073] S1. Obtain each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first cerebral hemorrhage image after registration with the second cerebral hemorrhage image. The first cerebral hemorrhage image is the cerebral hemorrhage image before follow-up, and the second cerebral hemorrhage image is the cerebral hemorrhage image after follow-up.

[0074] S2. Obtain each second hemorrhage lesion and the second location region of the second hemorrhage lesion from the second cerebral hemorrhage image;

[0075] S3. Determine the patient's cerebral hemorrhage type based on the first location region, the second location region, the first hemorrhage lesion, and the second hemorrhage lesion. The cerebral hemorrhage type includes at least one of the following: expanding or decreasing hemorrhage type, newly added hemorrhage type, and disappearing hemorrhage type.

[0076] S4. Determine a display layer based on the type of cerebral hemorrhage, and display the first hemorrhage lesion and the second hemorrhage lesion on the display layer.

[0077] In summary, the method provided in this disclosure obtains each first hemorrhage lesion and a first location region of the first hemorrhage lesion from a first hemorrhage image registered with a second hemorrhage image, wherein the first hemorrhage image is a pre-follow-up hemorrhage image and the second hemorrhage image is a post-follow-up hemorrhage image; obtains each second hemorrhage lesion and a second location region of the second hemorrhage lesion from the second hemorrhage image; determines the patient's hemorrhage type based on the first location region, the second location region, the first hemorrhage lesion, and the second hemorrhage lesion, wherein the hemorrhage type includes at least one of expanding or decreasing hemorrhage type, newly added hemorrhage type, and disappearing hemorrhage type; and determines a display layer based on the hemorrhage type to display the first hemorrhage lesion and the second hemorrhage lesion on the display layer. This method can determine all display layers of hemorrhage lesions according to the patient's hemorrhage type and display the first hemorrhage lesion and the second hemorrhage lesion on the display layer, enabling a complete presentation of the hemorrhage lesions to the doctor at once, thereby improving the doctor's diagnostic efficiency.

[0078] Below, in conjunction with specific embodiments, we will discuss... Figure 1 Each step in the process will be explained in detail:

[0079] In step S1, each first hemorrhage lesion and the first location region of the first hemorrhage lesion are obtained from the first brain hemorrhage image after being registered with the second brain hemorrhage image.

[0080] In one exemplary embodiment of this disclosure, the first brain hemorrhage image is a brain hemorrhage image before follow-up, and the second brain hemorrhage image is a brain hemorrhage image after follow-up.

[0081] In one exemplary embodiment of this disclosure, reference is made to Figure 2The system architecture shown may include: user-side mobile terminal device 201, user-side smart terminal device 204, and server 203. Data transmission between user-side mobile terminal device 201, user-side smart terminal device 204, and server 203 can all occur via network 202. The network may include various connection types, such as wired communication links, wireless communication links, etc. The aforementioned follow-up analysis method for cerebral hemorrhage can be executed on the server side, on the user-side terminal device, or by a collaboration between the user-side terminal device and the server side. Taking the above method as an example on the server side, the server first obtains a first brain hemorrhage image and a second brain hemorrhage image from the server. Then, after registering the first brain hemorrhage image and the second brain hemorrhage image, it obtains each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first brain hemorrhage image registered with the second brain hemorrhage image. The first brain hemorrhage image is the brain hemorrhage image before follow-up, and the second brain hemorrhage image is the brain hemorrhage image after follow-up. It then obtains each second hemorrhage lesion and the second location region of the second hemorrhage lesion from the second brain hemorrhage image. Based on the first location region, the second location region, the first hemorrhage lesion, and the second hemorrhage lesion, it determines the patient's brain hemorrhage type. The brain hemorrhage type includes at least one of the following: expanding or decreasing hemorrhage type, newly added hemorrhage type, and disappearing hemorrhage type. Then, it uses the brain hemorrhage type to determine the display layer, and then sends the display image to the user-side terminal device so that the user-side terminal device displays the first hemorrhage lesion and the second hemorrhage lesion on the display layer.

[0082] Based on the above, in an exemplary embodiment of this disclosure, obtaining each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first brain hemorrhage image registered with the second brain hemorrhage image includes:

[0083] S11. Normalize the registered first cerebral hemorrhage image to obtain a normalized cerebral hemorrhage image.

[0084] Based on the above, in one exemplary embodiment of this disclosure, the normalized brain hemorrhage image includes:

[0085] S111. The registered first cerebral hemorrhage image is normalized in size to obtain a normalized cerebral hemorrhage image.

[0086] S112. Determine the region of interest from the size-normalized brain hemorrhage image;

[0087] S113. Perform pixel normalization on the region of interest to obtain a pixel-normalized brain hemorrhage image.

[0088] In one exemplary embodiment of this disclosure, all brain hemorrhage images are three-dimensional images. To avoid the adverse effects of potentially outlier data, the registered first brain hemorrhage image is size-normalized to obtain a size-normalized brain hemorrhage image. Specifically, the slice thickness of the registered first brain hemorrhage image is isotropically normalized along the transverse, sagittal, and coronal directions. For example, the median slice thickness in the three directions is statistically analyzed, and the slice thickness in each direction is uniformly normalized to the median slice thickness in that direction. After the slice thickness of the registered first brain hemorrhage image is uniformly normalized to the median slice thickness in each direction, the median slice thickness in each direction is used to perform slice processing on the registered first brain hemorrhage image to obtain the slice processing result, which is the size-normalized brain hemorrhage image.

[0089] Furthermore, traditional image processing methods such as connected component extraction and thresholding are employed to remove the bed board and air from the size-normalized brain hemorrhage image, in order to extract the region of interest (ROI) from the size-normalized brain hemorrhage image. Connected component extraction involves first obtaining a binary image of the size-normalized brain hemorrhage image, then acquiring the region composed of connected pixels in this binary image, and finally using this region as the ROI. Thresholding segmentation identifies regions in the brain hemorrhage image where pixels fall within a preset pixel threshold range as the ROI.

[0090] Furthermore, after obtaining the region of interest (ROI), the pixels within the ROI are normalized to obtain a pixel-normalized image of brain hemorrhage. Specifically, a normal distribution is used to statistically analyze all voxels within the ROI. The voxels with the highest and lowest probabilities in the normal distribution curve of the ROI are removed. Then, the mean and variance var of all remaining voxels are obtained. These mean and variance are then used to normalize each voxel v within the ROI, resulting in the normalized voxel v = (v - mean) / var.

[0091] S12. Input the normalized cerebral hemorrhage image into a pre-trained hemorrhage lesion classification model to obtain the first hemorrhage lesion of the registered first cerebral hemorrhage image.

[0092] In an exemplary embodiment of this disclosure, the hemorrhage lesion classification model can be a three-dimensional fully convolutional neural network with a UNet structure. This hemorrhage lesion classification model can classify cerebral hemorrhage lesions into subarachnoid hemorrhage, parenchymal hemorrhage, intraventricular hemorrhage, subdural hemorrhage, and epidural hemorrhage. Therefore, after inputting the pixel-normalized cerebral hemorrhage image into the pre-trained hemorrhage lesion classification model, the first hemorrhage lesion can be obtained from the pixel-normalized cerebral hemorrhage image, and the probability of the first hemorrhage lesion belonging to subarachnoid hemorrhage, parenchymal hemorrhage, intraventricular hemorrhage, subdural hemorrhage, and epidural hemorrhage can be determined. Then, the probability of the first hemorrhage lesion belonging to subarachnoid hemorrhage, parenchymal hemorrhage, intraventricular hemorrhage, subdural hemorrhage, and epidural hemorrhage is calculated using argmax to determine the type of hemorrhage lesion of the first hemorrhage lesion.

[0093] S13. Determine the first location region of the first hemorrhage lesion in the registered first cerebral hemorrhage image.

[0094] In one exemplary embodiment of this disclosure, a three-dimensional coordinate system can be established on the registered first brain hemorrhage image, and then the three-dimensional coordinates of each voxel of the first hemorrhage lesion in the three-dimensional coordinate system can be determined. Then, the first location region of the first hemorrhage lesion in the registered first brain hemorrhage image can be determined based on the three-dimensional coordinates.

[0095] Based on the above, in an exemplary embodiment of this disclosure, before obtaining each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first cerebral hemorrhage image registered with the second cerebral hemorrhage image, the method further includes:

[0096] S41. Obtain a first training set, the first training set including a first brain hemorrhage image sample and a first sample label corresponding to the first brain hemorrhage image sample;

[0097] S42. Perform data augmentation on the first cerebral hemorrhage image sample to obtain an augmented first cerebral hemorrhage image sample.

[0098] S43. Input the augmented first cerebral hemorrhage image sample into the hemorrhage lesion classification model to obtain the first sample hemorrhage lesion type of the first cerebral hemorrhage image sample.

[0099] S44. Construct the first loss function of the bleeding lesion classification model using the bleeding lesion type of the first sample and the label;

[0100] S45. The first loss function is used to train the hemorrhage lesion classification model.

[0101] In one exemplary embodiment of this disclosure, the first cerebral hemorrhage image sample can be normalized in the same manner as steps S111 to S113 described above, and then the normalized first cerebral hemorrhage image sample can be augmented to obtain an augmented first cerebral hemorrhage image sample. Augmentation methods include random rotation, random noise, and gamma transformation, etc.

[0102] Furthermore, the augmented first cerebral hemorrhage image sample is hot-coded and input into the hemorrhage lesion classification model to obtain the first sample hemorrhage lesion type of the first cerebral hemorrhage image sample. Specifically, each voxel in the first cerebral hemorrhage image sample is output by the hemorrhage lesion classification model, and then after softmax, the output value of each channel represents the probability that the voxel belongs to the hemorrhage lesion type corresponding to that channel. Among them, the 0th channel corresponds to the background image, the 1st channel corresponds to subarachnoid hemorrhage, the 2nd channel corresponds to brain parenchymal hemorrhage, the 3rd channel corresponds to intraventricular hemorrhage, the 4th channel corresponds to subdural hemorrhage, and the 5th channel corresponds to epidural hemorrhage. Then, the first loss of the hemorrhage lesion classification model is directly calculated using the output obtained after softmax. In order to utilize the multi-scale features of the neural network and obtain richer semantic and deep information, for network structures containing different branches, the output prediction of the hemorrhage lesion classification model of the first cerebral hemorrhage image sample is combined with the network output prediction of the hemorrhage lesion classification model of different branches during training. i The first loss is calculated using the first sample label. In an exemplary embodiment of this disclosure, the first loss function can be constructed using methods such as cross-entropy, and then the first loss is calculated using the following formula:

[0103] loss1 = f(T1(predict) i .label1)).i=1.2...; (1)

[0104] Where loss1 represents the first loss, T1 represents the first loss function, label1 represents the first label, and f represents the calculation method such as weighted summation and average, which can be selected according to the actual situation.

[0105] For example, if the hemorrhage lesion classification model is a convolutional neural network with two output branches, and the outputs of the two output branches are predict1 and predict2, then the first loss calculated using a weighted summation and averaging method is as follows:

[0106]

[0107] Where a and b represent the weight parameters of the convolutional neural network for the two output branches, respectively.

[0108] In one exemplary embodiment of this disclosure, reasonable weight parameters can be set for subarachnoid hemorrhage, intracerebral hemorrhage, intraventricular hemorrhage, subdural hemorrhage, and epidural hemorrhage respectively. That is, each weight parameter is adjusted so that when the first loss is less than a preset threshold, the training of the hemorrhage lesion classification model is completed.

[0109] In step S2, each second hemorrhage lesion and the second location region of the second hemorrhage lesion are obtained from the second cerebral hemorrhage image.

[0110] In one exemplary embodiment of this disclosure, after training the hemorrhage lesion classification model using the above method, a second brain hemorrhage image can be input into the hemorrhage lesion classification model to obtain the second hemorrhage lesion. Similarly, a three-dimensional coordinate system can be established on the second brain hemorrhage image, which is consistent with the above coordinate system. Then, the three-dimensional coordinates of each voxel of the second hemorrhage lesion in the three-dimensional coordinate system are determined, and then the first location region of the second hemorrhage lesion in the second brain hemorrhage image is determined based on the three-dimensional coordinates.

[0111] In step S3, the type of cerebral hemorrhage in the patient is determined based on the first location region, the second location region, the first hemorrhage lesion, and the second hemorrhage lesion.

[0112] In one exemplary embodiment of this disclosure, the aforementioned types of cerebral hemorrhage include at least one of the following: hemorrhage types that expand or decrease, new hemorrhage types, and hemorrhage types that disappear.

[0113] Based on the above, in an exemplary embodiment of this disclosure, determining the type of cerebral hemorrhage based on the first location region, the second location region, the first hemorrhage lesion, and the second hemorrhage lesion includes:

[0114] S311. For each of the first bleeding lesions, if the bleeding lesion type of the first bleeding lesion is consistent with the bleeding lesion type of the second target bleeding lesion in the second bleeding lesion, and there is an overlapping area between the first location area and the second target location area of ​​the second target bleeding lesion, then obtain the first volume of the first location area and the second volume of the second location area.

[0115] S312. If the first volume is smaller than the second volume, then the type of cerebral hemorrhage is determined to be an expanding hemorrhage type.

[0116] S313. If the first volume is smaller than the second volume, then the type of cerebral hemorrhage is determined to be a reduced type of hemorrhage.

[0117] In one exemplary embodiment of this disclosure, the overlap between the first location region and the second target location region can be determined based on the three-dimensional coordinates of each first voxel in the first hemorrhage lesion and the three-dimensional coordinates of each first voxel in the second hemorrhage lesion. Specifically, if the three-dimensional coordinates of each first voxel in the first hemorrhage lesion are consistent with the three-dimensional coordinates of each second voxel in the second hemorrhage lesion, then it is determined whether the first location region and the second target location region overlap.

[0118] Further, after determining that the first location region and the second target location region overlap, the number of voxels in the first voxel within the first hemorrhage lesion is obtained. Then, based on the number of voxels and the unit volume of the voxels, the first volume of the first location region is determined. The second volume of the second target region is then obtained in the same manner. If the first volume is smaller than the second volume, it indicates that the hemorrhage area has increased after follow-up, and the type of cerebral hemorrhage is determined to be an expanding hemorrhage; if the first volume is smaller than the second volume, it indicates that the hemorrhage area has decreased after follow-up, and the type of cerebral hemorrhage is determined to be a decreasing hemorrhage.

[0119] Based on the above, in one exemplary embodiment of this disclosure, the method further includes:

[0120] S314. If the bleeding lesion type of the first bleeding lesion is inconsistent with the bleeding lesion type of each of the second bleeding lesions, then the bleeding type is determined to be the disappearance bleeding type.

[0121] In an exemplary embodiment of this disclosure, if the bleeding lesion type of the first bleeding lesion is inconsistent with the bleeding lesion type of each of the second bleeding lesions, it indicates that the bleeding lesion type of the first bleeding lesion is small after follow-up, and the bleeding type is determined to be the disappearance bleeding type.

[0122] Based on the above, in one exemplary embodiment of this disclosure, the method further includes:

[0123] S315. If a second specific bleeding lesion exists in the second bleeding lesion, then the bleeding type is determined to be a newly added bleeding type, and the bleeding lesion type of the second specific bleeding lesion is inconsistent with the bleeding lesion type of each of the first bleeding lesions.

[0124] In one exemplary embodiment of this disclosure, if the second bleeding lesion contains a second specific bleeding lesion whose bleeding lesion type is inconsistent with the bleeding lesion types of each of the first bleeding lesions, then the bleeding type is determined to be a newly added bleeding type. This indicates that a new type of bleeding lesion appears after follow-up, and the bleeding type is determined to be a newly added bleeding type.

[0125] It should be noted here that if the bleeding lesion type of the first bleeding lesion is the same as the bleeding lesion type of the second target bleeding lesion in the second bleeding lesion, but there is no overlap between the first location area and the second target location area of ​​the second target bleeding lesion, this indicates that the bleeding lesion disappeared before the follow-up and reappeared in a new location after the follow-up, and the bleeding type is also determined to be a newly added bleeding type.

[0126] In step S4, the display layer is determined using the type of cerebral hemorrhage, so as to display the first hemorrhage lesion and the second hemorrhage lesion on the display layer.

[0127] Based on the above, in an exemplary embodiment of this disclosure, determining the display layer using the type of cerebral hemorrhage includes:

[0128] S411. If the bleeding type is an expanding or decreasing bleeding type, then obtain the union region of the first location region and the second location region;

[0129] S412. Obtain the union region in the first layer of the first brain hemorrhage image / or the second brain hemorrhage image;

[0130] S413. Use the first layer as the display layer.

[0131] In an exemplary embodiment of this disclosure, if the hemorrhage type is an expanding or decreasing hemorrhage type, then the union region of the first location region and the second location region is obtained; then the first layer of the union region on the first brain hemorrhage image and / or the second brain hemorrhage image is obtained, the first layer being a layer where the union region is not empty in the z-axis direction on the first brain hemorrhage image and / or the second brain hemorrhage image; then the first layer of the first brain hemorrhage image and the first layer of the second brain hemorrhage image are displayed simultaneously, and the first location region of the first hemorrhage lesion in the first brain hemorrhage image is specially marked, and the second location region of the second hemorrhage lesion in the first brain hemorrhage image is specially marked.

[0132] Based on the above, in an exemplary embodiment of this disclosure, determining the display layer based on the type of cerebral hemorrhage includes:

[0133] S421. If the bleeding type is a newly added bleeding type, then obtain the second layer of the second location region in the second brain hemorrhage image;

[0134] S422. Use the second layer as the display layer.

[0135] In an exemplary embodiment of this disclosure, if the bleeding type is a newly added bleeding type, then the second layer of the second location region in the second brain hemorrhage image is obtained. The second layer is a layer in the second brain hemorrhage image where the second location region is not empty in the z-axis direction. Then, the second layer of the first brain hemorrhage image and the second layer of the first brain hemorrhage image are displayed simultaneously, and the second layer of the second brain hemorrhage image of the second location region of the second hemorrhage lesion is specially marked.

[0136] Based on the above, in an exemplary embodiment of this disclosure, determining the display layer based on the type of cerebral hemorrhage includes:

[0137] S431. If the bleeding type is a disappearing bleeding type, then obtain the third layer of the first location region in the first brain hemorrhage image.

[0138] S432. Use the third layer as the display layer.

[0139] In an exemplary embodiment of this disclosure, if the bleeding type is a disappearing bleeding type, then the third layer of the first location region in the first brain hemorrhage image is obtained. The third layer is a layer in the first brain hemorrhage image where the first location region is not empty in the z-axis direction. Then, the third layer of the first brain hemorrhage image and the third layer of the first brain hemorrhage image are displayed simultaneously, and the first location region of the first hemorrhage lesion is specially marked in the third layer of the first brain hemorrhage image.

[0140] Based on the above, in an exemplary embodiment of this disclosure, before obtaining each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first cerebral hemorrhage image registered with the second cerebral hemorrhage image, the method further includes:

[0141] S51. Obtain the deformation field of the first cerebral hemorrhage image relative to the second cerebral hemorrhage image;

[0142] In one exemplary embodiment of this disclosure, obtaining the deformation field of the first brain hemorrhage image relative to the second brain hemorrhage image includes:

[0143] S511. Obtain the first brain tissue from the first brain hemorrhage image;

[0144] In an exemplary embodiment of this disclosure, the first brain tissue includes a first midline of the brain, and the first brain tissue for acquiring the first brain hemorrhage image includes:

[0145] The first brain hemorrhage image is input into a pre-trained brain tissue extraction model to obtain an image of the first falx cerebri region in the first brain hemorrhage image; the middle pixel of each row of pixels in the first falx cerebri region image is obtained; and the line formed by the middle pixels is taken as the first brain midline.

[0146] In an exemplary embodiment of this disclosure, the brain tissue extraction model can be a UNet-type network. Before inputting the first brain hemorrhage image into the pre-trained brain tissue extraction model, a second training set is obtained, which includes second brain hemorrhage image samples and corresponding falx cerebri labels. The second brain hemorrhage image samples are augmented to obtain augmented second brain hemorrhage image samples. The augmented second brain hemorrhage image samples are input into the brain tissue extraction model to obtain the falx cerebri region of the second brain hemorrhage image samples. Then, a second loss function of the brain tissue extraction model is constructed based on the falx cerebri region and falx cerebri labels, and then the second loss of the brain tissue extraction model is calculated according to the second loss function. Specifically, the second loss is as follows:

[0147] loss2=f(T2(predict, label2)); (3)

[0148] Where Loss2 represents the second loss, T2 represents the second loss function, label2 represents the falx corpuscular label, and predict represents the falx corpuscular region.

[0149] Furthermore, the brain tissue extraction model is optimized using this second loss so that the second loss is less than a preset threshold, at which point the brain tissue extraction model training is complete.

[0150] In one exemplary embodiment of this disclosure, the brain tissue extraction model outputs two channels, corresponding to the background channel and the falx cerebri segmentation channel, respectively. Each voxel in the first brain hemorrhage image is output from the brain tissue extraction model, and after a softmax operation, the output value of the first channel represents the probability that the voxel belongs to the image background, and the output value of the second channel represents the probability that the voxel belongs to the falx cerebri.

[0151] Then, the argmax function is used to calculate the probability of each voxel in the first cerebral hemorrhage image belonging to the image background and the probability of belonging to the falx cerebri, thus obtaining the first falx cerebri region image of the first cerebral hemorrhage image. Further, the first falx cerebri region image is traversed row by row, and the pixel at the middle position of each row is taken, with the line formed by the middle pixel being used as the first midline of the brain.

[0152] In one exemplary embodiment of this disclosure, the first brain tissue further includes a first brain tissue outline. The brain tissue outline refers to the edge portion where the intracranial dura mater is located. In one exemplary embodiment of this disclosure, the brain tissue extraction model can also be trained using augmented second brain hemorrhage image samples and corresponding brain tissue structure labels. The training process is similar to that in the above embodiments, and will not be described again here.

[0153] Furthermore, the first brain hemorrhage image is input into a trained brain tissue extraction model to obtain the first brain tissue from the first brain hemorrhage image. Then, based on the first epidural hemorrhage region obtained from the hemorrhage lesion classification model, the remaining areas of the first brain tissue excluding the first epidural hemorrhage region are filled with holes to obtain the hole-filling result. Then, an edge extraction algorithm is used to extract the contour of the first brain tissue from the hole-filling result.

[0154] S512, Obtain the second brain tissue from the second brain hemorrhage image.

[0155] Similarly, the second brain tissue includes the second brain midline and the outline of the second brain tissue in the image of the second brain hemorrhage. The method for obtaining the second brain midline and the outline of the second brain tissue can be referred to the above embodiment, and will not be repeated here.

[0156] S513. Register the first brain hemorrhage image and the second brain hemorrhage image using the first brain tissue and the second brain tissue to obtain the deformation field;

[0157] S52. The first cerebral hemorrhage image is corrected using the deformation field to obtain the registered first cerebral hemorrhage image.

[0158] Currently, image registration algorithms are often used to align follow-up images of brain hemorrhage, marking lesions at the same location as matched lesions. However, when hemorrhage causes significant deformation of the internal brain structures, traditional registration methods struggle to align identical structures within the brain tissue. Furthermore, the falx cerebri and dura mater are unique soft tissue structures within the cranium; severe hemorrhage can cause deformation, but not severe structural damage. Therefore, aligning the falx cerebri and dura mater during registration, using this as reference information for brain tissue alignment, can improve the accuracy of matching lesions at the same location.

[0159] Specifically, the pixels at the first brain midline and the second brain tissue contour in the first brain hemorrhage image are labeled with different high signal values ​​that clearly distinguish brain tissue and skull pixels to obtain imageold_2. Similarly, the pixels at the second brain midline and the second brain tissue contour in the second brain hemorrhage image are labeled with different high signal values ​​that clearly distinguish brain tissue and skull pixels to obtain imagenew_2. Then, imageold_2 and imagenew_2 are registered using a non-rigid registration method to obtain a deformation field. This deformation field is then used to correct the first brain hemorrhage image, resulting in the registered first brain hemorrhage image.

[0160] In one exemplary embodiment of this disclosure, a non-rigid registration network can be used to register imageold_2 and imagenew_2. The non-rigid registration network can be a fully convolutional network, not limited to UNet. The input of the non-rigid registration network is the image pair to be matched (i.e., imageold_2 and imagenew_2), and the output is three three-channel images of the same size as the input images, each containing parameters dx, dy, and dz, representing the offsets of imageold_2 relative to imagenew_2 along the three axes, i.e., the deformation fields of imageold_2 relative to imagenew_2. Then, the pixel coordinates of imageold_2 are superimposed with the coordinate offsets and interpolated to shift the pixels of imageold_2 to the correct positions, thus completing the registration of the first cerebral hemorrhage image. During training, the non-rigid registration network uses normalized cross-correlation (NCC) similarity loss. Assuming the registered image pair is image_fixed and image_moved, the NCC loss is applied to measure the similarity between image_fixed and image_moved with superimposed deformation field. L2 regularization is applied to the deformation field output by the network to make the registered image smoother.

[0161] In summary, the follow-up analysis method for cerebral hemorrhage provided in this disclosure can analyze the changes in lesions on a category-by-category and lesion-by-lesion basis, including the expansion, reduction, disappearance, and new addition of hemorrhage; it can also extract the brain midline and brain contour from cerebral hemorrhage images, and perform registration analysis on the cerebral hemorrhage images based on the brain midline and brain contour, which can improve the accuracy of brain tissue structure matching and avoid follow-up lesion matching errors caused by poor registration results due to significant changes in brain tissue structure caused by severe hemorrhage; in addition, it can also present the hemorrhage lesion to the doctor in a complete manner at one time, thereby improving the doctor's diagnostic efficiency.

[0162] After introducing the follow-up analysis method for cerebral hemorrhage according to exemplary embodiments of the present invention, the following will refer to... Figure 3An exemplary embodiment of the cerebral hemorrhage follow-up analysis device of the present invention will be described.

[0163] refer to Figure 3 As shown, the cerebral hemorrhage follow-up analysis device 30 of an exemplary embodiment of the present invention may include:

[0164] The first hemorrhage lesion acquisition module 301 is used to acquire each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first brain hemorrhage image after registration with the second brain hemorrhage image. The first brain hemorrhage image is the brain hemorrhage image before follow-up, and the second brain hemorrhage image is the brain hemorrhage image after follow-up.

[0165] The second hemorrhage lesion acquisition module 302 is used to acquire each second hemorrhage lesion and the second location region of the second hemorrhage lesion from the second brain hemorrhage image;

[0166] The cerebral hemorrhage type determination module 303 is used to determine the type of cerebral hemorrhage of a patient based on the first location region, the second location region, the first hemorrhage lesion, and the second hemorrhage lesion. The type of cerebral hemorrhage includes at least one of the following: expanding or decreasing hemorrhage type, newly added hemorrhage type, and disappearing hemorrhage type.

[0167] Display layer determination module 304 is used to determine a display layer based on the type of cerebral hemorrhage, so as to display the first hemorrhage lesion and the second hemorrhage lesion on the display layer.

[0168] In one exemplary embodiment of this disclosure, the cerebral hemorrhage type determination module includes:

[0169] The first volume acquisition unit is used to acquire the first volume of the first location region and the second volume of the second location region for each first bleeding lesion if the bleeding lesion type of the first bleeding lesion is consistent with the bleeding lesion type of the second target bleeding lesion in the second bleeding lesion, and there is an overlapping area between the first location region and the second target location region of the second target bleeding lesion.

[0170] An expansion hemorrhage type determination unit is used to determine that the cerebral hemorrhage type is an expansion hemorrhage type if the first volume is smaller than the second volume;

[0171] The reduced hemorrhage type determination unit is used to determine that the cerebral hemorrhage type is a reduced hemorrhage type if the first volume is smaller than the second volume.

[0172] A disappearance hemorrhage type determination unit, configured in an exemplary embodiment of this disclosure, wherein the brain hemorrhage type determination module further includes:

[0173] If the bleeding lesion type of the first bleeding lesion is inconsistent with the bleeding lesion type of each of the second bleeding lesions, then the bleeding type is determined to be the disappearance bleeding type.

[0174] In one exemplary embodiment of this disclosure, the cerebral hemorrhage type determination module further includes:

[0175] A new bleeding type determination unit is added, which is used to determine the bleeding type as a new bleeding type if a second specific bleeding lesion exists in the second bleeding lesion, and the bleeding lesion type of the second specific bleeding lesion is inconsistent with the bleeding lesion type of each of the first bleeding lesions.

[0176] In one exemplary embodiment of this disclosure, the display layer determination module includes:

[0177] The union region acquisition unit is used to acquire the union region of the first location region and the second location region if the bleeding type is an expanding or decreasing bleeding type.

[0178] The first layer acquisition unit is used to acquire the union region in the first layer of the first brain hemorrhage image / or the second brain hemorrhage image;

[0179] The first display layer determining unit is used to determine the first layer as the display layer.

[0180] In one exemplary embodiment of this disclosure, the display layer determination module includes:

[0181] The second layer acquisition unit is used to acquire the second layer of the second location region in the second brain hemorrhage image if the hemorrhage type is a newly added hemorrhage type.

[0182] The second layer acquisition unit is used to use the second layer as the display layer.

[0183] In one exemplary embodiment of this disclosure, the display layer determination module includes:

[0184] The third layer acquisition unit is used to acquire the third layer of the first location region in the first brain hemorrhage image if the hemorrhage type is the disappearing hemorrhage type.

[0185] The third layer acquisition unit is used to use the third layer as the display layer.

[0186] In one exemplary embodiment of this disclosure, the apparatus further includes:

[0187] Image registration module, the image registration module comprising:

[0188] Deformation field acquisition unit, used to acquire the deformation field of the first cerebral hemorrhage image relative to the second cerebral hemorrhage image;

[0189] An image registration unit is used to correct the first cerebral hemorrhage image using the deformation field to obtain a registered first cerebral hemorrhage image.

[0190] In one exemplary embodiment of this disclosure, the deformation field acquisition unit includes:

[0191] The first brain tissue acquisition unit is used to acquire the first brain tissue from the first brain hemorrhage image.

[0192] The second brain tissue acquisition unit is used to acquire the second brain tissue from the second brain hemorrhage image.

[0193] The deformation field acquisition subunit is used to register the first brain hemorrhage image and the second brain hemorrhage image using the first brain tissue and the second brain tissue to acquire the deformation field.

[0194] In one exemplary embodiment of this disclosure, the first brain tissue includes a first midline of the brain, and the first brain tissue acquisition unit includes:

[0195] The first falx cerebri region image acquisition unit is used to input the first cerebral hemorrhage image into a pre-trained brain tissue extraction model to obtain the first falx cerebri region image in the first cerebral hemorrhage image.

[0196] The intermediate pixel acquisition unit is used to acquire the intermediate pixel of each row of pixels in the first falx cerebralis region image;

[0197] The first brain midline acquisition subunit is used to take the line formed by the intermediate pixels as the first brain midline.

[0198] In one exemplary embodiment of this disclosure, the first hemorrhage lesion acquisition module includes:

[0199] The image normalization unit is used to normalize the registered first cerebral hemorrhage image to obtain a normalized cerebral hemorrhage image.

[0200] The first hemorrhage lesion acquisition unit is used to input the pixel-normalized cerebral hemorrhage image into a pre-trained hemorrhage lesion classification model to obtain the first hemorrhage lesion of the registered first cerebral hemorrhage image.

[0201] The first location region determination unit is used to determine the first location region of the first hemorrhagic lesion in the registered first cerebral hemorrhage image.

[0202] In one exemplary embodiment of this disclosure, the image normalization unit includes:

[0203] A size normalization unit is used to normalize the size of the registered first cerebral hemorrhage image to obtain a size-normalized cerebral hemorrhage image.

[0204] A region of interest determination unit is used to determine the region of interest from the size-normalized brain hemorrhage image;

[0205] A pixel normalization unit is used to normalize the pixels of the region of interest to obtain a pixel-normalized brain hemorrhage image.

[0206] In one exemplary embodiment of this disclosure, the apparatus further includes:

[0207] Model training unit, used for

[0208] A training set acquisition unit is used to acquire a training set, which includes brain hemorrhage image samples and sample labels corresponding to the brain hemorrhage image samples.

[0209] The sample augmentation unit is used to augment the brain hemorrhage image sample to obtain an augmented brain hemorrhage image sample.

[0210] The sample hemorrhage lesion type acquisition unit is used to input the brain hemorrhage image sample into the hemorrhage lesion classification model to obtain the sample hemorrhage lesion type of the brain hemorrhage image sample;

[0211] The loss function construction unit is used to construct the loss function of the bleeding lesion classification model using the sample bleeding lesion type and the label;

[0212] The bleeding lesion classification model training unit is used to train the bleeding lesion classification model using the loss function.

[0213] Since the functional modules of the cerebral hemorrhage follow-up analysis device in this invention are the same as those in the above-described cerebral hemorrhage follow-up analysis method, they will not be described again here.

[0214] After introducing the cerebral hemorrhage follow-up analysis method and cerebral hemorrhage follow-up analysis device according to exemplary embodiments of the present invention, the following will refer to... Figure 4 A storage medium according to an exemplary embodiment of the present invention will be described. (See reference...) Figure 4 As shown, a program product 400 for implementing the above-described method according to an embodiment of the present invention is described. This product may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a device such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0215] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0216] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting programs for use by or in conjunction with an instruction execution system, apparatus, or device.

[0217] The program code contained on the readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, fiber optic, RF, etc., or any suitable combination thereof. The program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0218] After introducing the storage medium of exemplary embodiments of the present invention, the following references are made. Figure 5 An electronic device according to an exemplary embodiment of the present invention will be described.

[0219] Figure 5 The electronic device 50 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0220] like Figure 5As shown, the electronic device 50 is manifested in the form of a general-purpose computing device. The components of the electronic device 50 may include, but are not limited to: at least one processing unit 510, at least one storage unit 520, a bus 530 connecting different system components (including the storage unit 520 and the processing unit 510), and a display unit 540. The storage unit stores program code that can be executed by the processing unit 510, causing the processing unit 510 to perform the steps described in the "Exemplary Methods" section above, according to various exemplary embodiments of the present invention. For example, the processing unit 510 may perform, as... Figure 1 Steps S1 to S4 are shown in the diagram.

[0221] Storage unit 520 may include volatile storage units, such as random access memory (RAM) 5201 and / or cache memory 5202, and may further include read-only memory (ROM) 5203. Storage unit 520 may also include a program / utility 5204 having a set (at least one) of program modules 5205, such program modules 5205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0222] Bus 530 may include a data bus, an address bus, and a control bus.

[0223] Electronic device 50 can also communicate with one or more external devices 60 (e.g., keyboard, pointing device, Bluetooth device, etc.) via input / output (I / O) interface 550. Electronic device 50 also includes a display unit 540 connected to input / output (I / O) interface 550 for display purposes. Furthermore, electronic device 50 can communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 560. As shown, network adapter 560 communicates with other modules of electronic device 50 via bus 530. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 50, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems. It should be noted that although several modules or sub-modules of the cerebral hemorrhage follow-up analysis device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the present invention, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0224] Furthermore, although the operations of the method of the present invention are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all of the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0225] While the spirit and principles of the invention have been described with reference to several specific embodiments, it should be understood that the invention is not limited to the disclosed specific embodiments, and the division of aspects does not imply that features in these aspects cannot be combined for benefit; such division is merely for ease of description. The invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

Claims

1. A method for follow-up analysis of cerebral hemorrhage, characterized in that, include: Each first hemorrhage lesion and the first location region of the first hemorrhage lesion are obtained from the first brain hemorrhage image after registration with the second brain hemorrhage image. The first brain hemorrhage image is the brain hemorrhage image before follow-up, and the second brain hemorrhage image is the brain hemorrhage image after follow-up. Obtain each second hemorrhage lesion and the second location region of the second hemorrhage lesion from the second cerebral hemorrhage image; The patient's cerebral hemorrhage type is determined based on the first location region, the second location region, the first hemorrhage lesion, and the second hemorrhage lesion. The cerebral hemorrhage type includes at least one of the following: expanding or decreasing hemorrhage type, newly added hemorrhage type, and disappearing hemorrhage type. The type of cerebral hemorrhage is used to determine a display layer, in which the first hemorrhage lesion and the second hemorrhage lesion are displayed.

2. The method according to claim 1, characterized in that, The determination of the patient's cerebral hemorrhage type based on the first location region, the second location region, the first hemorrhage lesion, and the second hemorrhage lesion includes: For each of the first hemorrhage lesions, if the hemorrhage lesion type of the first hemorrhage lesion is consistent with the hemorrhage lesion type of the second target hemorrhage lesion in the second hemorrhage lesion, and there is an overlapping area between the first location region and the second target location region of the second target hemorrhage lesion, then the first volume of the first location region and the second volume of the second location region are obtained. If the first volume is smaller than the second volume, then the type of cerebral hemorrhage is determined to be an expanding hemorrhage type; If the first volume is smaller than the second volume, then the type of cerebral hemorrhage is determined to be a reduced type of hemorrhage.

3. The method according to claim 2, characterized in that, The method further includes: If the bleeding lesion type of the first bleeding lesion is inconsistent with the bleeding lesion type of each of the second bleeding lesions, then the bleeding type is determined to be the disappearance bleeding type.

4. The method according to claim 2, characterized in that, The method further includes: If a second specific bleeding lesion exists within the second bleeding lesion, then the bleeding type is determined to be a newly added bleeding type, and the bleeding lesion type of the second specific bleeding lesion is inconsistent with the bleeding lesion type of each of the first bleeding lesions.

5. The method according to claim 1, characterized in that, The step of determining the display layer based on the type of cerebral hemorrhage includes: If the bleeding type is an expanding or decreasing bleeding type, then obtain the union region of the first location region and the second location region; Obtain the union region in the first layer of the first brain hemorrhage image / or the second brain hemorrhage image; The first layer is used as the display layer.

6. The method according to claim 1, characterized in that, The determination of the display layer based on the type of cerebral hemorrhage includes: If the hemorrhage type is a newly added hemorrhage type, then obtain the second location region in the second layer of the second brain hemorrhage image; The second layer is used as the display layer.

7. The method according to claim 1, characterized in that, The determination of the display layer based on the type of cerebral hemorrhage includes: If the bleeding type is a disappearing bleeding type, then obtain the third layer of the first location region in the first brain hemorrhage image; The third layer is used as the display layer.

8. The method according to claim 1, characterized in that, Before obtaining each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first brain hemorrhage image registered with the second brain hemorrhage image, the method further includes: Obtain the deformation field of the first brain hemorrhage image relative to the second brain hemorrhage image; The deformation field is used to correct the first cerebral hemorrhage image to obtain the registered first cerebral hemorrhage image.

9. The method according to claim 8, characterized in that, The process of acquiring the deformation field of the first brain hemorrhage image relative to the second brain hemorrhage image includes: Obtain the first brain tissue from the first brain hemorrhage image; Second brain tissue was obtained from the second brain hemorrhage image; The deformation field is obtained by registering the first brain hemorrhage image and the second brain hemorrhage image using the first brain tissue and the second brain tissue.

10. The method according to claim 9, characterized in that, The first brain tissue includes a first midline of the brain, and the first brain tissue from which the first brain hemorrhage image is acquired includes: The first brain hemorrhage image is input into a pre-trained brain tissue extraction model to obtain an image of the first falx cerebri region in the first brain hemorrhage image. Obtain the middle pixel of each row of pixels in the first falx cerebralis region image; The line formed by the intermediate pixels is taken as the first midline of the brain.

11. The method according to claim 1, characterized in that, The step of obtaining each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first brain hemorrhage image registered with the second brain hemorrhage image includes: The registered first cerebral hemorrhage image is normalized to obtain a normalized cerebral hemorrhage image; The normalized cerebral hemorrhage image is input into a pre-trained hemorrhage lesion classification model to obtain the first hemorrhage lesion of the registered first cerebral hemorrhage image; The first location region of the first hemorrhage lesion in the registered first cerebral hemorrhage image is determined.

12. The method according to claim 11, characterized in that, The normalization of the registered first cerebral hemorrhage image to obtain a normalized cerebral hemorrhage image includes: The registered first cerebral hemorrhage image is normalized in size to obtain a normalized cerebral hemorrhage image; Determine the region of interest from the size-normalized brain hemorrhage image; Pixel normalization is performed on the region of interest to obtain a pixel-normalized brain hemorrhage image.

13. The method according to claim 11, characterized in that, Before obtaining each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first brain hemorrhage image registered with the second brain hemorrhage image, the method further includes: Obtain a training set, which includes brain hemorrhage image samples and sample labels corresponding to the brain hemorrhage image samples; The brain hemorrhage image samples are augmented to obtain augmented brain hemorrhage image samples; The brain hemorrhage image sample is input into the hemorrhage lesion classification model to obtain the sample hemorrhage lesion type of the brain hemorrhage image sample; The loss function of the hemorrhage lesion classification model is constructed using the sample hemorrhage lesion type and the label; The loss function is used to train the hemorrhage lesion classification model.

14. A follow-up analysis device for cerebral hemorrhage, characterized in that, include: The first hemorrhage lesion acquisition module is used to acquire each first hemorrhage lesion and the first location region of the first hemorrhage lesion from the first brain hemorrhage image after registration with the second brain hemorrhage image. The first brain hemorrhage image is the brain hemorrhage image before follow-up, and the second brain hemorrhage image is the brain hemorrhage image after follow-up. The second hemorrhage lesion acquisition module is used to acquire each second hemorrhage lesion and the second location region of the second hemorrhage lesion from the second brain hemorrhage image; A brain hemorrhage type determination module is used to determine the type of brain hemorrhage of a patient based on the first location region, the second location region, the first hemorrhage lesion, and the second hemorrhage lesion. The type of brain hemorrhage includes at least one of the following: expanding or decreasing hemorrhage type, newly added hemorrhage type, and disappearing hemorrhage type. The display layer determination module is used to determine a display layer based on the type of cerebral hemorrhage, so as to display the first hemorrhage lesion and the second hemorrhage lesion on the display layer.

15. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the brain hemorrhage follow-up analysis method as described in any one of claims 1 to 13.

Images